from brenier to kntohe and from knothe to brenier
play

From Brenier to Kntohe and from Knothe to Brenier: convergence, PDE - PowerPoint PPT Presentation

Apr 11, 2024 •231 likes •704 views

Monotone transports Convergence Semi-discrete evolution Continuous evolution From Brenier to Kntohe and from Knothe to Brenier: convergence, PDE and numerical ideas Filippo Santambrogio Laboratoire de Math ematiques dOrsay, Universit

  1. Monotone transports Convergence Semi-discrete evolution Continuous evolution From Brenier to Kntohe and from Knothe to Brenier: convergence, PDE and numerical ideas Filippo Santambrogio Laboratoire de Math´ ematiques d’Orsay, Universit´ e Paris-Sud http://www.math.u-psud.fr/ ∼ santambr/ Grenoble, October 4th, 2013, Modelisation with optimal transport logo Filippo Santambrogio From Brenier to Knothe, from Knothe to Brenier

  2. Monotone transports Convergence Semi-discrete evolution Continuous evolution Monotone trasports 1 The 1D monotone transport The Brenier map, gradient of a convex potential The Knothe-Rosenblatt map Convergence as t → 0 for the cost | x 1 − y 1 | 2 + t | x 2 − y 2 | 2 2 A conjecture by Y. Brenier A proof in the spirit of Γ − developments Assumptions and counter-examples Atoms in the disintegrated measures Dynamics as t moves in the semi-discrete case 3 An ODE for the potential Evolution of cells Dynamics in the continuous case 4 The PDE for the potential The initial condition Well-posedness Numerical solution logo Filippo Santambrogio From Brenier to Knothe, from Knothe to Brenier

  3. Monotone transports Convergence Semi-discrete evolution Continuous evolution Monotone trasports 1 The 1D monotone transport The Brenier map, gradient of a convex potential The Knothe-Rosenblatt map Convergence as t → 0 for the cost | x 1 − y 1 | 2 + t | x 2 − y 2 | 2 2 A conjecture by Y. Brenier A proof in the spirit of Γ − developments Assumptions and counter-examples Atoms in the disintegrated measures Dynamics as t moves in the semi-discrete case 3 An ODE for the potential Evolution of cells Dynamics in the continuous case 4 The PDE for the potential The initial condition Well-posedness Numerical solution logo Filippo Santambrogio From Brenier to Knothe, from Knothe to Brenier

  4. Monotone transports Convergence Semi-discrete evolution Continuous evolution Monotone trasports 1 The 1D monotone transport The Brenier map, gradient of a convex potential The Knothe-Rosenblatt map Convergence as t → 0 for the cost | x 1 − y 1 | 2 + t | x 2 − y 2 | 2 2 A conjecture by Y. Brenier A proof in the spirit of Γ − developments Assumptions and counter-examples Atoms in the disintegrated measures Dynamics as t moves in the semi-discrete case 3 An ODE for the potential Evolution of cells Dynamics in the continuous case 4 The PDE for the potential The initial condition Well-posedness Numerical solution logo Filippo Santambrogio From Brenier to Knothe, from Knothe to Brenier

  5. Monotone transports Convergence Semi-discrete evolution Continuous evolution Monotone trasports 1 The 1D monotone transport The Brenier map, gradient of a convex potential The Knothe-Rosenblatt map Convergence as t → 0 for the cost | x 1 − y 1 | 2 + t | x 2 − y 2 | 2 2 A conjecture by Y. Brenier A proof in the spirit of Γ − developments Assumptions and counter-examples Atoms in the disintegrated measures Dynamics as t moves in the semi-discrete case 3 An ODE for the potential Evolution of cells Dynamics in the continuous case 4 The PDE for the potential The initial condition Well-posedness Numerical solution logo Filippo Santambrogio From Brenier to Knothe, from Knothe to Brenier

  6. Monotone transports Convergence Semi-discrete evolution Continuous evolution Monotone transports 1D, Brenier, Knothe logo Filippo Santambrogio From Brenier to Knothe, from Knothe to Brenier

  7. Monotone transports Convergence Semi-discrete evolution Continuous evolution Very briefly, something you all know about the optimal transport problem � Monge Problem : min c ( x , T ( x )) µ ( dx ) : T # µ = ν proposed by G. Monge in 1781, for c ( x , y ) = | x − y | . � Kantorovich Problem : (1942) min c ( x , y ) d γ : γ ∈ Π( µ, ν ) where Π( µ, ν ) := { γ : ( π x ) ♯ γ = µ, ( π y ) ♯ γ = ν } . This gives again Monge’s framework when γ = ( id × T ) # µ . Advantages of Kantorovich’s formulation it’s a convex problem it always has a solution (if c is l.s.c.) il has a dual formulation : � � � min c d γ = sup φ d µ + ψ d ν : φ ( x ) + ψ ( y ) ≤ c ( x , y ) . logo Filippo Santambrogio From Brenier to Knothe, from Knothe to Brenier

  8. Monotone transports Convergence Semi-discrete evolution Continuous evolution Very briefly, something you all know about the optimal transport problem � Monge Problem : min c ( x , T ( x )) µ ( dx ) : T # µ = ν proposed by G. Monge in 1781, for c ( x , y ) = | x − y | . � Kantorovich Problem : (1942) min c ( x , y ) d γ : γ ∈ Π( µ, ν ) where Π( µ, ν ) := { γ : ( π x ) ♯ γ = µ, ( π y ) ♯ γ = ν } . This gives again Monge’s framework when γ = ( id × T ) # µ . Advantages of Kantorovich’s formulation it’s a convex problem it always has a solution (if c is l.s.c.) il has a dual formulation : � � � min c d γ = sup φ d µ + ψ d ν : φ ( x ) + ψ ( y ) ≤ c ( x , y ) . logo Filippo Santambrogio From Brenier to Knothe, from Knothe to Brenier

  9. Monotone transports Convergence Semi-discrete evolution Continuous evolution The monotone transport in 1D Given µ, ν ∈ P ( R ), if µ has no atoms, there exists unique an increasing map T : R → R such that T # µ = ν . If F and G are the cumulative distribution functions of µ and ν , respec- tively, and if G is strictly increasing on spt ν (i.e. if spt ν is an interval), we can compute it through T = G − 1 ◦ F (if ν has not full support a generalized inverse of G should be used). This map turns out to be optimal for all the costs of the form c ( x , y ) = h ( x − y ) with h convex (and it is the unique optimizer if h is strictly convex). In particular, this covers the quadratic case c ( x , y ) = | x − y | 2 . It is very easy to compute, but only works in 1D. logo Filippo Santambrogio From Brenier to Knothe, from Knothe to Brenier

  10. Monotone transports Convergence Semi-discrete evolution Continuous evolution The monotone transport in 1D Given µ, ν ∈ P ( R ), if µ has no atoms, there exists unique an increasing map T : R → R such that T # µ = ν . If F and G are the cumulative distribution functions of µ and ν , respec- tively, and if G is strictly increasing on spt ν (i.e. if spt ν is an interval), we can compute it through T = G − 1 ◦ F (if ν has not full support a generalized inverse of G should be used). This map turns out to be optimal for all the costs of the form c ( x , y ) = h ( x − y ) with h convex (and it is the unique optimizer if h is strictly convex). In particular, this covers the quadratic case c ( x , y ) = | x − y | 2 . It is very easy to compute, but only works in 1D. logo Filippo Santambrogio From Brenier to Knothe, from Knothe to Brenier

  11. Monotone transports Convergence Semi-discrete evolution Continuous evolution The monotone transport in 1D Given µ, ν ∈ P ( R ), if µ has no atoms, there exists unique an increasing map T : R → R such that T # µ = ν . If F and G are the cumulative distribution functions of µ and ν , respec- tively, and if G is strictly increasing on spt ν (i.e. if spt ν is an interval), we can compute it through T = G − 1 ◦ F (if ν has not full support a generalized inverse of G should be used). This map turns out to be optimal for all the costs of the form c ( x , y ) = h ( x − y ) with h convex (and it is the unique optimizer if h is strictly convex). In particular, this covers the quadratic case c ( x , y ) = | x − y | 2 . It is very easy to compute, but only works in 1D. logo Filippo Santambrogio From Brenier to Knothe, from Knothe to Brenier

  12. Monotone transports Convergence Semi-discrete evolution Continuous evolution The quadratic cost in R d If d > 1 and µ, ν are measures on Ω ⊂ R d the situation is trickier but Brenier proved the following : if µ is nice (for instance µ ≪ L d ), then there exists unique an optimal map, and it given by T = ∇ φ , with φ convex. It has some monotonicity property (for instance, DT is a symmetric and positive definite matrix). But it is trickier to compute. The change-of- variable-formula, if µ = f ( x ) dx and ν = g ( y ) dy , gives the Jacobian f condition det DT = g ◦ T , which reads here f det( D 2 φ ) = g ◦ ∇ φ, with φ convex , (Monge-Amp` ere equation). Its “boundary” condition is given by ∇ φ ( x ) ∈ Ω for all x ∈ Ω. This PDE is nonlinear and difficult to solve, both nume- rically and theoretically. Some regularity theorems exist giving φ ∈ C k +2 ,α if f , g are bounded from below and belong to C k ,α and spt ν is convex.. In this case T is C k +1 ,α . Y. Brenier , Polar factorization and monotone rearrangement of vector-valued logo functions, CPAM , 1991. Filippo Santambrogio From Brenier to Knothe, from Knothe to Brenier

More recommend